A PV Prediction Model Based on Sparrow Search Optimization with Variational Mode Decomposition and Gated Recurrent Unit Neural Network

In order to address the issue of inaccurate prediction due to the intermittent and fluctuating nature of photovoltaic output, this study puts forward a model for short-term photovoltaic power prediction. This model is based on variational mode decomposition (VMD) and utilizes a gated recurrent unit (GRU) neural network, which has been optimized using the Sparrow Search algorithm (SSA). As a first step, in order to select model inputs that are strongly correlated with PV power, Pearson correlation coefficient (PCC) was used. Secondly, using SSA to optimize VMD and GRU parameters respectively, combined with decomposed historical PV power data and highly correlated historical meteorological factor data, PV forecast power is obtained. Finally, the proposed model and GRU and VMD-GRU results are evaluated with four error indices. The findings indicate that the suggested approach significantly enhances the forecast precision of solar power.